• Media type: E-Article
  • Title: EXERCISE OF MACHINE LEARNING USING SOME PYTHON TOOLS AND TECHNIQUES
  • Contributor: Geldiev, Ertan Mustafa; Nenkov, Nayden Valkov; Petrova, Mariana Mateeva
  • Published: CBU Research Institute, 2018
  • Published in: CBU International Conference Proceedings, 6 (2018), Seite 1062-1070
  • Language: Without Specification
  • DOI: 10.12955/cbup.v6.1295
  • ISSN: 1805-9961; 1805-997X
  • Keywords: General Medicine
  • Origination:
  • Footnote:
  • Description: One of the goals of predictive analytics training using Python tools is to create a "Model" from classified examples that classifies new examples from a Dataset. The purpose of different strategies and experiments is to create a more accurate prediction model. The goals we set out in the study are to achieve successive steps to find an accurate model for a dataset and preserving it for its subsequent use using the python instruments. Once we have found the right model, we save it and load it later, to classify if we have "phishing" in our case. In the case that the path we reach to the discovery of the search model, we can ask ourselves how much we can automate everything and whether a computer program can be written to automatically go through the unified steps and to find the right model? Due to the fact that the steps for finding the exact model are often unified and repetitive for different types of data, we have offered a hypothetical algorithm that could write a complex computer program searching for a model, for example when we have a classification task. This algorithm is rather directional and does not claim to be all-encompassing. The research explores some features of Python Scientific Python Packages like Numpy, Pandas, Matplotlib, Scipy and scycit-learn to create a more accurate model. The Dataset used for the research was downloaded free from the UCI Machine Learning Repository (UCI Machine Learning Repository, 2017).